Mintlify's Admin MCP Can Now Comment, Suggest and Skip Pull Requests

Mintlify has updated its Admin MCP server so connected AI agents can leave comments and suggestions on documentation, and edits no longer have to go through a pull request. According to the company, any MCP-compatible agent, such as Claude Code, Cursor, Codex or ChatGPT, can now edit pages, flag changes for review and publish small fixes straight to main.
- Mintlify's Admin MCP server can now leave comments and suggestions on documentation.
- MCP edits no longer require a pull request, and agents can publish small fixes straight to main.
- Branches now commit automatically and appear in the web editor for team review.
- Mintlify says Claude Code connected to the MCP scores roughly the same as its editor agent on internal evals.
- Agents connect via https://mcp.mintlify.com after signing in with a Mintlify account.
No more waiting on a pull request
Until now, every change made through the Mintlify MCP went through a git branch and a pull request. An agent could draft an update, but a person still had to find the PR and merge it before anything shipped, according to Mintlify.
That has changed. Branches now commit automatically and show up in the web editor, so the work lands in the place where a team already reviews changes. For small fixes, the agent can publish directly to main without the extra step.
The update was announced on Mintlify's blog on October 19, 2026, in a post by David Isquick of Product Marketing.
Comments and suggestions instead of rewrites
The other addition is that agents can now propose changes rather than rewriting pages outright. A person can then accept or reject the suggestion. Agents can also leave comments to flag something for a team member to look at.
Mintlify says these additions mean the MCP now covers everything the Mintlify agent does. The company also reports that, on its internal evals, Claude Code connected to the MCP scores roughly the same as the editor agent. Its conclusion is that users can work with whichever tool they already have open. These are the company's own evaluation results, and the post does not give figures.
Mintlify also says it optimized every tool call for speed and token usage, so that long editing sessions cost less and finish sooner. No specific numbers were given for these improvements.
How to get started
Setup starts by adding https://mcp.mintlify.com as an MCP server in a compatible client. The user then signs in with a Mintlify account when prompted.
From there, the agent can be given a task in plain language. Mintlify's example is: "Update the authentication guide to reflect our new token expiry and leave a comment for the API team to review." That single request combines an edit with a comment aimed at a specific team.
Mintlify points users to its MCP documentation for setup instructions by client and the full list of tools.
What it means for documentation teams
For teams that already use AI agents to maintain docs, the change removes a manual step that sat between a drafted edit and a published page. Because branches appear in the web editor, review can happen in the same place as other documentation work, and the suggestion and comment features give reviewers a way to stay in control of larger changes.
The post also lists two other recent Mintlify releases: a Savings section in Mintlify Analytics that shows the time and money teams save with automations and the assistant, and editor performance work that the company says lets very large pages open, navigate and edit up to 100x faster.
Mintlify: Improved Admin MCP


